Episodes (Page 6)
✨
Dharmesh Shah (HubSpot co-founder) introduces 'hybrid teams' concept
Dharmesh Shah
✨
Snipd evolved from promising design to best-in-class podcast app with AI search/transcript features after Overcast 2.0 disappointed with no AI improvements in 3 years
✨
OpenAI released Operator and Deep Research agents in first 2 months of 2025, now bringing agent capabilities to API via Responses API, Web Search, Computer Use, and File Search tools
✨
Anthropic's Claude 3.5 Sonnet (Sonnet 3.7) successfully plays Pokémon Red live on Twitch with custom harness enabling screen vision, navigation, and game state memory
David Hershey
✨
Paul Klein (Browserbase founder) discusses building browser infrastructure specifically optimized for AI agents, including handling CAPTCHAs, geolocation, and multimodal inputs at scale
✨
Deep Research represents a new category combining agentic workflows with custom-tuned frontier models (o3, Gemini 1.5 Flash), moving beyond basic LLM-powered search
✨
Bee AI, an always-on wearable ambient agent founded by Maria and Ethan, emerged as the biggest winner from 2024's failed AI wearables wave (Rabbit, Humane, Friend.com)
✨
Bret Taylor discusses the emergence of 'AI Architect' as new role at Sierra, with leaders responsible for defining, managing, and scaling conversational AI/CX platforms across enterprise
Bret Taylor
✨
Samuel Colvin (Pydantic founder) raised $17M from Sequoia to commercialize Pydantic, one of top 25 PyPI packages with 300M+ monthly downloads used by OpenAI SDK and FastAPI
Samuel Colvin
✨
Karina Nguyen, Research Manager at OpenAI, discusses new agent reasoning interfaces including o1/o3, Claude 3, ChatGPT Canvas, Tasks, and Operator for AI-assisted work
✨
Chai Research (founded by former hedge fund traders) hit 1.4M DAU and >$22M revenue, outlasting Noam Shazeer's Character AI through superior engineering and crowdsourced model development
William Beauchamp
✨
DeepSeek v3 launched as best open-weights model (LM Arena score 1319), making large model serving the bottleneck, requiring specialized inference infrastructure like Baseten's H200 clusters
✨
Simon Willison, top AI blogger, recaps 2024 learnings about LLMs in crossover episode with TechMeme Ride Home's Brian, targeting tech-literate non-technical audiences
✨
Will Bryk (Exa.ai) building neural PageRank alternative with $5M H200 investment to move search from links-based to embeddings-based context retrieval for LLMs
✨
ComfyUI founder comfyanonymous discusses open source AI image generation, building alternative to AUTOMATIC1111 after involvement with Stability in 2023
✨
Latent Space celebrates 100th episode milestone with year-in-review covering rise of AI engineering, competitive landscape evolution, and major conferences (LIVE events, NeurIPS, ICLR, ICML)
✨
Latent Space LIVE! at NeurIPS 2024 keynote on state of LLM agents features Professor Graham Neubig covering 2024 agent developments and industry relevance
✨
2024 dominated by synthetic data rise, with major shift from real data-only training paradigm toward hybrid and fully synthetic approaches
✨
Post-transformer architectures (State Space Models, RWKV) gaining traction as researchers hunt for next 'Attention Is All You Need' breakthrough beyond scaling transformers
✨
Open models exploded in 2024 with shift from five dominant players (2023) to diverse ecosystem of Mistral, Mosaic, and others competing in top LLM ranks